Powering Private AI: VMware by Broadcom + AMD Instinct MI350 GPUs

At VMware Explore 2025 (Day Three), hosts Pete Flecha and John Nicholson sit down with Raj Bhat from AMD and Dave Morera from Broadcom to discuss a major announcement: VMware’s support for AMD Instinct MI350 GPUs for Private AI workloads.

 

Raj dives into the specs of the MI350—288GB of memory and nearly 20 petaflops of performance, packaged in a UBB form factor with eight GPUs connected together—making it ideal for running large AI models. Dave explains how this partnership expands customer hardware choices while delivering the same seamless Private AI software experience powered by VMware Cloud Foundation.

 

The conversation also highlights:

• How AMD’s ROCM (Radeon Open Compute Model) framework enables flexibility with AI models.

• VMware Cloud Foundation’s automation for deploying these GPUs.

• AMD’s open-source and free software approach, reducing subscription costs.

• Hardware availability through OEM partners like Supermicro and Dell, with both air-cooled and water-cooled options.

• Compatibility with both AMD and Intel CPUs.

 

This session showcases how VMware and AMD are working together to give customers more choice, efficiency, and scalability in Private AI.